@cursor/july 0.1.93 → 0.1.95

package/dist/docs/assets/reference_project-layout.md.WN9nwJht.js

@@ -1,17 +0,0 @@
-import{_ as t,c as a,o as s,ag as d}from"./chunks/framework.BCISBCiQ.js";const g=JSON.parse('{"title":"Project layout","description":"The folder structure under agent/ and the path-derived naming rule.","frontmatter":{"title":"Project layout","description":"The folder structure under agent/ and the path-derived naming rule."},"headers":[],"relativePath":"reference/project-layout.md","filePath":"reference/project-layout.md"}'),o={name:"reference/project-layout.md"};function n(r,e,i,c,l,h){return s(),a("div",null,[...e[0]||(e[0]=[d(`<h1 id="project-layout" tabindex="-1">Project layout <a class="header-anchor" href="#project-layout" aria-label="Permalink to &quot;Project layout&quot;">​</a></h1><p>The Agent SDK builds an agent by walking the filesystem under <code>agent/</code>. Each folder has a defined purpose. The path a file lands in determines how the Agent SDK loads it.</p><h2 id="folder-structure" tabindex="-1">Folder structure <a class="header-anchor" href="#folder-structure" aria-label="Permalink to &quot;Folder structure&quot;">​</a></h2><p>For the capabilities below, identity usually comes from the path. A/B experiments can override their file-derived name.</p><table tabindex="0"><thead><tr><th>Path</th><th>Resolves to</th></tr></thead><tbody><tr><td><code>agent/tools/approve_pr.ts</code></td><td>tool <code>approve_pr</code></td></tr><tr><td><code>agent/mcp-connections/linear.ts</code></td><td>MCP connection <code>linear</code></td></tr><tr><td><code>agent/skills/pr-review.md</code></td><td>skill <code>pr-review</code></td></tr><tr><td><code>agent/subagents/reviewer/</code></td><td>subagent <code>reviewer</code></td></tr><tr><td><code>agent/channels/drive.ts</code></td><td>channel <code>drive</code>, routes under <code>/v1/channels/drive</code></td></tr><tr><td><code>agent/ab.ts</code></td><td>A/B experiment <code>ab</code> unless <code>name</code> overrides it</td></tr><tr><td><code>agent/ab/concise.ts</code></td><td>A/B experiment <code>concise</code> unless <code>name</code> overrides it</td></tr></tbody></table><p>The root agent takes its name from <code>package.json</code> <code>name</code>, falling back to the directory name. When serving multiple agents, the slug is the directory name and must match <code>[A-Za-z0-9][A-Za-z0-9_-]*</code> (and not the reserved <code>v1</code>, <code>playground</code>, or <code>docs</code> segments).</p><h2 id="project-overview" tabindex="-1">Project overview <a class="header-anchor" href="#project-overview" aria-label="Permalink to &quot;Project overview&quot;">​</a></h2><p>Most projects start with this shape.</p><div class="language-text vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">text</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span>my-agent/</span></span>
-<span class="line"><span>├── package.json</span></span>
-<span class="line"><span>├── agent/</span></span>
-<span class="line"><span>│   ├── agent.ts               # runtime config (model, runtime, cloud/local)</span></span>
-<span class="line"><span>│   ├── instructions.md        # always-on system prompt (required)</span></span>
-<span class="line"><span>│   ├── tools/</span></span>
-<span class="line"><span>│   │   └── approve_pr.ts      # one typed tool per file</span></span>
-<span class="line"><span>│   ├── skills/</span></span>
-<span class="line"><span>│   │   └── pr-review.md       # on-demand procedures (SKILL.md convention)</span></span>
-<span class="line"><span>│   ├── mcp-connections/</span></span>
-<span class="line"><span>│   │   └── linear.ts          # tools from external MCP servers</span></span>
-<span class="line"><span>│   └── channels/</span></span>
-<span class="line"><span>│       └── github.ts          # messages and external events</span></span>
-<span class="line"><span>└── evals/</span></span>
-<span class="line"><span>    └── readiness.eval.ts      # regression cases</span></span></code></pre></div><p>Evals live in <code>evals/</code> at the project root, a sibling of <code>agent/</code>, never inside it. <code>agent/evals/</code> is silently ignored. See <a href="./../evals.html">Evals</a>.</p><h2 id="folder-reference" tabindex="-1">Folder reference <a class="header-anchor" href="#folder-reference" aria-label="Permalink to &quot;Folder reference&quot;">​</a></h2><p>Each path maps to a capability and a reference page.</p><table tabindex="0"><thead><tr><th>Path</th><th>What it is</th><th>Reference</th></tr></thead><tbody><tr><td><code>agent/agent.ts</code></td><td><code>defineAgent({ model?, runtime?, cloud?, local? })</code>; the model defaults to <code>grok-4.5</code> with <code>effort=high</code>, <code>fast=true</code></td><td><a href="./agent-config.html">Agent config</a></td></tr><tr><td><code>agent/instructions.md</code></td><td>Always-on system prompt, required on the root agent (<code>.ts</code> and directory forms exist)</td><td><a href="./instructions.html">Instructions</a></td></tr><tr><td><code>agent/tools/&lt;name&gt;.ts</code></td><td>One typed tool; filename = tool name. <code>execution: &quot;server&quot;</code> (in-process, default) or <code>&quot;agent&quot;</code> (a script that runs where the agent runs)</td><td><a href="./tools.html">Tools</a></td></tr><tr><td><code>agent/skills/*</code></td><td>SKILL.md-convention procedures, loaded on demand</td><td><a href="./skills.html">Skills</a></td></tr><tr><td><code>agent/mcp-connections/&lt;name&gt;.ts</code></td><td>MCP servers, available to the model, to server tools (<code>ctx.host.mcp</code>), and to channel/schedule handlers (<code>args.host.mcp</code>)</td><td><a href="./connections.html">MCP connections</a></td></tr><tr><td><code>agent/subagents/&lt;id&gt;/</code></td><td>Child agent directory; <code>description</code> required</td><td><a href="./subagents.html">Subagents</a></td></tr><tr><td><code>agent/channels/*.ts</code></td><td>HTTP surfaces beyond the built-in session API; <code>slack.ts</code> and <code>github.ts</code> use the platform packs</td><td><a href="./channels.html">Channels</a></td></tr><tr><td><code>agent/hooks/*.ts</code></td><td>Observe-only event subscribers, never fatal</td><td><a href="./hooks.html">Hooks</a></td></tr><tr><td><code>agent/otel.ts</code></td><td><code>defineOtel</code> OTLP export (traces, metrics, optional logs)</td><td><a href="./../guides/opentelemetry.html">OpenTelemetry</a></td></tr><tr><td><code>agent/ab.ts</code>, <code>agent/ab/*.ts</code></td><td><code>defineAB</code> experiments with sticky variants and live metrics</td><td><a href="./../ab.html">Live A/B metrics</a></td></tr><tr><td><code>agent/ab.config.ts</code></td><td><code>defineABConfig</code> shared A/B settings</td><td><a href="./../ab.html">Live A/B metrics</a></td></tr><tr><td><code>agent/storage.ts</code></td><td><code>defineStorage</code> backend for the durable <code>host.kv</code> / <code>host.files</code> APIs</td><td><a href="./../storage.html">Storage</a></td></tr><tr><td><code>agent/artifacts.ts</code></td><td><code>defineArtifacts</code> kinds, the <code>tag_artifact</code> opt-in, and retention</td><td><a href="./artifacts.html">Artifacts</a></td></tr><tr><td><code>agent/schedules/*</code></td><td>Cron-driven runs (UTC, 5-field; never auto-fire under <code>--dev</code>)</td><td><a href="./schedules.html">Schedules</a></td></tr><tr><td><code>agent/sandbox/workspace/**</code></td><td>Seed files copied into each local session workspace</td><td><a href="./sessions.html#what-goes-into-a-local-session-workspace">Sessions</a></td></tr><tr><td><code>agent/playground/</code></td><td>Custom playground tool chips</td><td><a href="./playground.html">Playground</a></td></tr><tr><td><code>agent/lib/</code></td><td>Import-only shared code, never discovered</td><td>None</td></tr><tr><td><code>evals/evals.config.ts</code></td><td>Shared eval settings (e.g. <code>maxConcurrency</code>); required when evals exist</td><td><a href="./../evals.html">Evals</a></td></tr><tr><td><code>evals/**/*.eval.ts</code></td><td>Filesystem evals; case id = path under <code>evals/</code></td><td><a href="./../evals.html">Evals</a></td></tr></tbody></table><p><code>agent/lib/</code> is the only place for shared code. Everything else under <code>agent/</code> is discovery surface. A stray <code>.ts</code> file in one of these folders is treated as a definition.</p><h2 id="why-didn-t-the-agent-sdk-discover-my-file" tabindex="-1">Why didn&#39;t the Agent SDK discover my file? <a class="header-anchor" href="#why-didn-t-the-agent-sdk-discover-my-file" aria-label="Permalink to &quot;Why didn&#39;t the Agent SDK discover my file?&quot;">​</a></h2><p>Run <code>agent-sdk validate --dir .</code> and <code>agent-sdk info --dir .</code>. <code>validate</code> prints diagnostics, and <code>serve</code> refuses to start on error-severity ones. Warnings, such as cloud runtime combined with local-only capabilities, print but don&#39;t block. <code>info</code> lists the discovered surface, so a missing tool or channel shows up immediately. From there, check the folder reference: the file is usually in the wrong directory or has the wrong extension.</p><div class="language-bash vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">bash</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">agent-sdk</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> validate</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> --dir</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> .</span><span style="--shiki-light:#6A737D;--shiki-dark:#6A737D;">    # diagnostics; non-zero exit on errors</span></span>
-<span class="line"><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">agent-sdk</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> info</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> --dir</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> .</span><span style="--shiki-light:#6A737D;--shiki-dark:#6A737D;">        # human-readable surface</span></span>
-<span class="line"><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">agent-sdk</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> info</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> --dir</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> .</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> --json</span><span style="--shiki-light:#6A737D;--shiki-dark:#6A737D;"> # machine-readable project info (same shape as GET /v1/info)</span></span></code></pre></div><h2 id="what-s-next" tabindex="-1">What&#39;s next <a class="header-anchor" href="#what-s-next" aria-label="Permalink to &quot;What&#39;s next&quot;">​</a></h2><p>Continue with these pages:</p><ul><li><a href="./agent-config.html">Agent config</a>: the runtime config at the root</li><li><a href="./tools.html">Tools</a>: add typed actions under <code>agent/tools/</code></li><li><a href="./../ab.html">Live A/B metrics</a>: compare variants from <code>agent/ab.ts</code> or <code>agent/ab/</code></li><li><a href="./../concepts.html">Concepts</a>: why the filesystem is the interface</li></ul>`,20)])])}const u=t(o,[["render",n]]);export{g as __pageData,u as default};

package/dist/docs/assets/troubleshooting.md.vCWwvqcJ.js

@@ -1 +0,0 @@
-import{_ as t,c as o,o as a,ag as d}from"./chunks/framework.BCISBCiQ.js";const p=JSON.parse('{"title":"Fix common agent problems","description":"Match what you see to a cause, then read a session trace when you need more detail.","frontmatter":{"title":"Fix common agent problems","description":"Match what you see to a cause, then read a session trace when you need more detail."},"headers":[],"relativePath":"troubleshooting.md","filePath":"troubleshooting.md"}'),r={name:"troubleshooting.md"};function s(n,e,c,i,l,h){return a(),o("div",null,[...e[0]||(e[0]=[d('<h1 id="fix-common-agent-problems" tabindex="-1">Fix common agent problems <a class="header-anchor" href="#fix-common-agent-problems" aria-label="Permalink to &quot;Fix common agent problems&quot;">​</a></h1><p>Start with four checks, in order:</p><ol><li>Project discovery: <code>agent-sdk validate --dir .</code></li><li>Whether the serve process is running</li><li>What the playground or HTTP API shows</li><li>The session event stream (trace)</li></ol><p>Match your symptom below. Keep the commands as <code>agent-sdk</code>. If it is not on <code>PATH</code>, use <code>npx @cursor/july</code>.</p><h2 id="what-if-serve-or-the-playground-looks-wrong" tabindex="-1">What if serve or the playground looks wrong? <a class="header-anchor" href="#what-if-serve-or-the-playground-looks-wrong" aria-label="Permalink to &quot;What if serve or the playground looks wrong?&quot;">​</a></h2><table tabindex="0"><thead><tr><th>What you see</th><th>What to do</th></tr></thead><tbody><tr><td><code>serve</code> won&#39;t start</td><td>Run <code>agent-sdk validate --dir .</code> and fix the reported errors.</td></tr><tr><td>Playground is blank or says there are no agents</td><td>The UI needs a running <code>serve</code> process. Building the playground assets alone is not enough.</td></tr><tr><td>The playground UI looks stale</td><td><code>serve --dev</code> prints a playground URL. Open that URL. Agent-file edits still need a restart (press Enter on the TTY).</td></tr><tr><td>Sessions exist but the playground list is empty</td><td>The list shows sessions for the authenticated caller. In <code>--dev</code> on loopback the list is wider. Otherwise open <code>/&lt;slug&gt;/playground?sessionId=ses_…</code>.</td></tr><tr><td>Port 3000 is already in use</td><td>For the default serve port, the CLI tries the next free port and prints a notice. Pass <code>--port</code> to pick one, or <code>--port 0</code> for any free port. Stop leftover playground or webhook-forwarder processes if you need the original port.</td></tr></tbody></table><h2 id="what-if-a-model-turn-goes-wrong" tabindex="-1">What if a model turn goes wrong? <a class="header-anchor" href="#what-if-a-model-turn-goes-wrong" aria-label="Permalink to &quot;What if a model turn goes wrong?&quot;">​</a></h2><table tabindex="0"><thead><tr><th>What you see</th><th>What to do</th></tr></thead><tbody><tr><td>Built-in file reads and greps fail; the turn retries for a long time</td><td>Run under Node 22.13+ (or <code>tsx</code>), never Bun. Look for <code>NGHTTP2_FRAME_SIZE_ERROR</code> in logs.</td></tr><tr><td>The turn fails immediately with an API-key error</td><td>Sign in with <code>agent-sdk login</code>, or set <code>CURSOR_API_KEY</code>. Discovery, <code>info</code>, <code>call</code>, and serve bring-up work without a key; model turns need one.</td></tr><tr><td>Replies quote rules or <code>AGENTS.md</code> from outside your agent project</td><td>The session workspace inherited parent-folder config. Nested git checkouts default <code>local.cwd</code> to a per-project cache directory under <code>~/.cache</code>. Point <code>defineAgent({ local: { cwd } })</code> at a checkout only when the agent should inherit that tree, or set <code>--state-root</code> to a clean directory (for example under <code>/tmp</code>).</td></tr><tr><td>Yellow box shows Datadog/Linear tools, but the model lists <code>GetDynamicTools</code> / IDE <code>cursor</code> tools and never calls them</td><td>Attached MCP sits behind harness meta-tools, or <code>hostOnly</code> hid the connection, or the harness cwd is still inside another checkout. Set <code>advertiseTools: true</code> for named tools on local turns. Check <code>GET /v1/info</code> <code>local.cwd</code> and <code>connections[].advertiseTools</code>.</td></tr><tr><td>Server tools, skills, or workspace seed files never appear</td><td>Server tools and sandbox seeds apply on the local runtime (cloud server tools need <code>--public-url</code> / <code>--cloud-tools-url</code>). Skills reach cloud through the Agent Store when hosting or a personal <code>CURSOR_API_KEY</code> is available; otherwise only skills already in the cloud repo. <code>validate</code> warns when this combination is present.</td></tr><tr><td><code>validate</code> and <code>run</code> succeed, but typecheck fails in CI</td><td>The CLI runs TypeScript with type-stripping only. Keep tool <code>execute</code> return types as object literals or <code>type</code> aliases, not <code>interface</code> types.</td></tr><tr><td>Login works, but turns are rejected when using custom API hosts</td><td>Point login and model traffic at the same host (<code>CURSOR_API_BASE_URL</code> and <code>CURSOR_BACKEND_URL</code>). A key from one host is rejected by the other.</td></tr></tbody></table><h2 id="what-if-the-http-api-returns-an-error" tabindex="-1">What if the HTTP API returns an error? <a class="header-anchor" href="#what-if-the-http-api-returns-an-error" aria-label="Permalink to &quot;What if the HTTP API returns an error?&quot;">​</a></h2><table tabindex="0"><thead><tr><th>What you see</th><th>What to do</th></tr></thead><tbody><tr><td><code>409</code> on a follow-up message</td><td>Refresh the <code>continuationToken</code> or confirm the session is a chat session. Task sessions do not accept follow-ups.</td></tr><tr><td><code>409 session_busy</code> on <code>call --session</code></td><td>Wait for the model turn to finish, or omit <code>--session</code> for a one-off call.</td></tr><tr><td><code>403</code> on stream or follow-up</td><td>Use the same auth identity that created the session. Off localhost, pass <code>--bearer-token</code> and send it on every request.</td></tr><tr><td>Works on localhost; blocked through a tunnel or LAN</td><td>Default auth allows only direct loopback callers. Share the host with <code>--bearer-token &lt;secret&gt;</code> (or authored <code>bearerAuth</code>). Use <code>--allow-anonymous</code> only on a trusted private network.</td></tr><tr><td>A channel route fails to compile with a schema type error</td><td><code>GET</code> routes need a Zod <code>querySchema</code>. <code>POST</code> / <code>PUT</code> / <code>PATCH</code> need a Zod <code>bodySchema</code>. Use <code>z.object({})</code> or <code>z.unknown()</code> for open shapes.</td></tr></tbody></table><h2 id="what-if-github-webhooks-misbehave" tabindex="-1">What if GitHub webhooks misbehave? <a class="header-anchor" href="#what-if-github-webhooks-misbehave" aria-label="Permalink to &quot;What if GitHub webhooks misbehave?&quot;">​</a></h2><table tabindex="0"><thead><tr><th>What you see</th><th>What to do</th></tr></thead><tbody><tr><td><code>github forward</code> returns 401 on every delivery, but the hook was created</td><td>Clear <code>GITHUB_TOKEN</code> and <code>GH_TOKEN</code> for that command. The forwarder uses your <code>gh</code> CLI login: <code>GITHUB_TOKEN= GH_TOKEN= agent-sdk github forward …</code></td></tr><tr><td><code>Hook already exists</code> when starting a forwarder</td><td>GitHub allows one forwarder per repo. Run a single <code>github forward --dir &lt;parent&gt;</code> and stop stale forwarders.</td></tr><tr><td>Deliveries rejected outside <code>--dev</code></td><td>Set <code>GITHUB_WEBHOOK_SECRET</code> on the server and on the signer. Without a secret, the channel stays loopback-only.</td></tr><tr><td>You lack repo admin and can&#39;t forward</td><td>Use <code>agent-sdk github replay &lt;pr-url&gt;</code>. It needs pull access only and posts signed test payloads.</td></tr></tbody></table><h2 id="what-if-slack-stays-quiet" tabindex="-1">What if Slack stays quiet? <a class="header-anchor" href="#what-if-slack-stays-quiet" aria-label="Permalink to &quot;What if Slack stays quiet?&quot;">​</a></h2><table tabindex="0"><thead><tr><th>What you see</th><th>What to do</th></tr></thead><tbody><tr><td>Logs show <code>channel idle … missing credentials</code></td><td>Expected when tokens are missing. Run <code>agent-sdk slack create --dir &lt;agent&gt;</code> to provision the app and write the tokens, or <code>agent-sdk slack init --manual --dir &lt;agent&gt;</code> and paste the manifests at api.slack.com. Then set <code>&lt;PREFIX&gt;_SLACK_BOT_TOKEN</code> and <code>&lt;PREFIX&gt;_SLACK_APP_TOKEN</code> per agent and run <code>agent-sdk slack doctor --prefix &lt;PREFIX&gt;</code>.</td></tr><tr><td><code>slack create</code> reports the app needs admin approval</td><td>Open Slack&#39;s <strong>Request approval</strong> page (the CLI prints the link; the same URL is <strong>Send a reminder</strong> after you submit). Managed install does not file the request. Keep the CLI running, then click <strong>Retry</strong> in the dashboard after an admin approves.</td></tr><tr><td>The bot ignores ordinary channel posts</td><td>Default engagement is mentions and DMs only. Enable <code>engagement.channelPosts</code> with an allowlist, and subscribe the app to <code>message.channels</code> / <code>message.groups</code>.</td></tr><tr><td>Approve / Deny buttons do nothing</td><td>Channels that post approval cards need <code>toolApprovals: true</code>. Recreate the app with <code>slack create</code> if interactivity is off.</td></tr></tbody></table><h2 id="what-if-host-mcp-oauth-fails" tabindex="-1">What if host MCP OAuth fails? <a class="header-anchor" href="#what-if-host-mcp-oauth-fails" aria-label="Permalink to &quot;What if host MCP OAuth fails?&quot;">​</a></h2><table tabindex="0"><thead><tr><th>What you see</th><th>What to do</th></tr></thead><tbody><tr><td><code>must be defineConnection({ url, oauth: true })</code></td><td>The connection file needs <code>oauth: true</code>, or you passed the wrong connection name to <code>agent-sdk mcp oauth</code>.</td></tr><tr><td>Local auth works; hosted calls unauthorized</td><td>Run <code>agent-sdk mcp oauth &lt;name&gt; --store</code>, confirm names with <code>agent-sdk secrets list &lt;slug&gt;</code>, then redeploy.</td></tr><tr><td>Model asks for <code>mcp_auth</code> or IDE MCP for a privileged server</td><td>That connection is <code>hostOnly</code>. Call it from a host tool via <code>ctx.host.mcp</code>, and update instructions.</td></tr></tbody></table><p>See <a href="./guides/mcp-oauth.html">Host MCP OAuth</a> and <a href="./../skills/mcp-auth/SKILL.html"><code>skills/mcp-auth/SKILL.md</code></a>.</p><h2 id="what-if-a-secret-showed-up-in-a-terminal-transcript" tabindex="-1">What if a secret showed up in a terminal transcript? <a class="header-anchor" href="#what-if-a-secret-showed-up-in-a-terminal-transcript" aria-label="Permalink to &quot;What if a secret showed up in a terminal transcript?&quot;">​</a></h2><table tabindex="0"><thead><tr><th>What you see</th><th>What to do</th></tr></thead><tbody><tr><td><code>secrets set … NAME=VALUE</code> in an agent-captured terminal or shell history</td><td>Rotate the secret at the provider. Set it again with names only: <code>agent-sdk secrets set &lt;slug&gt; NAME</code> (hidden prompt) or pipe/redirect the value. <code>NAME=VALUE</code> requires <code>--from-argv</code> and still leaks into argv.</td></tr><tr><td>Alias token printed during first deploy or <code>rotate-token</code></td><td>Treat it as exposed if the transcript left your machine. Run <code>agent-sdk rotate-token &lt;slug&gt;</code>, store the new token outside agent transcripts, and update callers.</td></tr><tr><td>Someone verified a secret with <code>echo</code> / <code>printenv</code></td><td>Rotate it. Confirm presence with <code>agent-sdk secrets list &lt;slug&gt;</code> (names only), then redeploy and test the feature.</td></tr></tbody></table><h2 id="what-if-schedules-reminders-or-approvals-stall" tabindex="-1">What if schedules, reminders, or approvals stall? <a class="header-anchor" href="#what-if-schedules-reminders-or-approvals-stall" aria-label="Permalink to &quot;What if schedules, reminders, or approvals stall?&quot;">​</a></h2><table tabindex="0"><thead><tr><th>What you see</th><th>What to do</th></tr></thead><tbody><tr><td>A schedule or reminder never fires under <code>--dev</code></td><td>Dev mode does not auto-fire. Trigger with <code>POST /&lt;slug&gt;/v1/dev/schedules/&lt;id&gt;</code> or <code>POST /&lt;slug&gt;/v1/dev/reminders/&lt;id&gt;</code> (list reminders at <code>GET /v1/dev/reminders</code>).</td></tr><tr><td>A pending tool approval disappeared after restart</td><td>Parked approvals do not survive host restart. They resolve as interrupted. Run the turn again.</td></tr><tr><td>A reminder is disarmed after restart (<code>handler_lost_on_restart</code>)</td><td>Handler-form reminders live in memory. Re-arm them from the code that created them, or use prompt-form reminders.</td></tr></tbody></table><h2 id="how-do-i-read-a-session-trace" tabindex="-1">How do I read a session trace? <a class="header-anchor" href="#how-do-i-read-a-session-trace" aria-label="Permalink to &quot;How do I read a session trace?&quot;">​</a></h2><p>Look at <code>actions.requested</code> / <code>action.result</code> pairs for the tool trajectory. Count calls by tool name before blaming latency. Separate host-side work (channel <code>callTool</code>, preparation) from tools the model chose.</p><p><code>turn.failed</code> with <code>&quot;turn interrupted&quot;</code> means a follow-up or stop ended the turn on purpose.</p><p>If the model reads outside the session workspace, the prepared files don&#39;t match what the instructions expect. Fix the layout. See <a href="./hillclimbing.html">Hillclimbing</a>.</p><p><code>agent-sdk trajectory --events &lt;file&gt;</code> summarizes any saved NDJSON stream. The playground <strong>Open trace</strong> control does the same visually.</p><h2 id="what-s-next" tabindex="-1">What&#39;s next <a class="header-anchor" href="#what-s-next" aria-label="Permalink to &quot;What&#39;s next&quot;">​</a></h2><ul><li><a href="./concepts.html">Concepts</a>: the model behind these symptoms</li><li><a href="./hillclimbing.html">Hillclimbing</a>: when the agent runs but underperforms</li><li><a href="./deployment.html">Deployment</a>: auth and state on shared hosts</li></ul>',28)])])}const m=t(r,[["render",s]]);export{p as __pageData,m as default};

package/dist/docs/example-agents/approval-buddy.html

@@ -1,36 +0,0 @@
-<!DOCTYPE html>
-<html lang="en-US" dir="ltr">
-  <head>
-    <meta charset="utf-8">
-    <meta name="viewport" content="width=device-width,initial-scale=1">
-    <title>Keep PR approval policy deterministic with Approval Buddy | Agent SDK</title>
-    <meta name="description" content="Separate code-owned eligibility from model-owned review, then connect GitHub, Slack, subagents, durable storage, and evals.">
-    <meta name="generator" content="VitePress v1.6.4">
-    <link rel="preload stylesheet" href="/docs/assets/style.BRuM8477.css" as="style">
-    <link rel="preload stylesheet" href="/docs/vp-icons.css" as="style">
-    
-    <script type="module" src="/docs/assets/app.CjWU-x0z.js"></script>
-    <link rel="preload" href="/docs/assets/inter-roman-latin.Di8DUHzh.woff2" as="font" type="font/woff2" crossorigin="">
-    <link rel="modulepreload" href="/docs/assets/chunks/theme.Dvq1Bktu.js">
-    <link rel="modulepreload" href="/docs/assets/chunks/framework.BCISBCiQ.js">
-    <link rel="modulepreload" href="/docs/assets/example-agents_approval-buddy.md.DmezILPg.lean.js">
-    <script id="check-dark-mode">(()=>{const e=localStorage.getItem("vitepress-theme-appearance")||"auto",a=window.matchMedia("(prefers-color-scheme: dark)").matches;(!e||e==="auto"?a:e==="dark")&&document.documentElement.classList.add("dark")})();</script>
-    <script id="check-mac-os">document.documentElement.classList.toggle("mac",/Mac|iPhone|iPod|iPad/i.test(navigator.platform));</script>
-    <link rel="alternate" type="text/plain" href="/docs/llms.txt">
-    <link rel="alternate" type="text/markdown" href="/docs/example-agents/approval-buddy.md">
-  </head>
-  <body>
-    <div id="app"><div class="Layout" data-v-282c430e><!--[--><!--]--><!--[--><span tabindex="-1" data-v-af8643bc></span><a href="#VPContent" class="VPSkipLink visually-hidden" data-v-af8643bc>Skip to content</a><!--]--><!----><header class="VPNav" data-v-282c430e data-v-4751689a><div class="VPNavBar" data-v-4751689a data-v-d2a336f3><div class="wrapper" data-v-d2a336f3><div class="container" data-v-d2a336f3><div class="title" data-v-d2a336f3><div class="VPNavBarTitle has-sidebar" data-v-d2a336f3 data-v-3a787a7b><a class="title" href="/docs/" data-v-3a787a7b><!--[--><!--]--><!----><span data-v-3a787a7b>Agent SDK</span><!--[--><!--[--><!--[--><!--[--><span class="agent-sdk-version" title="@cursor/july 0.1.93" data-v-c26a5f4f>0.1.93</span><!--]--><!--]--><!--]--><!--]--></a></div></div><div class="content" data-v-d2a336f3><div class="content-body" data-v-d2a336f3><!--[--><!--]--><div class="VPNavBarSearch search" data-v-d2a336f3><!--[--><!----><div id="local-search"><button type="button" class="DocSearch DocSearch-Button" aria-label="Search"><span class="DocSearch-Button-Container"><span class="vp-icon DocSearch-Search-Icon"></span><span class="DocSearch-Button-Placeholder">Search</span></span><span class="DocSearch-Button-Keys"><kbd class="DocSearch-Button-Key"></kbd><kbd class="DocSearch-Button-Key">K</kbd></span></button></div><!--]--></div><nav aria-labelledby="main-nav-aria-label" class="VPNavBarMenu menu" data-v-d2a336f3 data-v-b60c0a58><span id="main-nav-aria-label" class="visually-hidden" data-v-b60c0a58> Main Navigation </span><!--[--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/quickstart.html" tabindex="0" data-v-b60c0a58 data-v-74de87b9><!--[--><span data-v-74de87b9>Quickstart</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/guides/webhooks.html" tabindex="0" data-v-b60c0a58 data-v-74de87b9><!--[--><span data-v-74de87b9>Guides</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/example-agents/" tabindex="0" data-v-b60c0a58 data-v-74de87b9><!--[--><span data-v-74de87b9>Examples</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/reference/project-layout.html" tabindex="0" data-v-b60c0a58 data-v-74de87b9><!--[--><span data-v-74de87b9>Reference</span><!--]--></a><!--]--><!--]--></nav><!----><div class="VPNavBarAppearance appearance" data-v-d2a336f3 data-v-d559fc20><button class="VPSwitch VPSwitchAppearance" type="button" role="switch" title aria-checked="false" data-v-d559fc20 data-v-deaafa36 data-v-88da49d2><span class="check" data-v-88da49d2><span class="icon" data-v-88da49d2><!--[--><span class="vpi-sun sun" data-v-deaafa36></span><span class="vpi-moon moon" data-v-deaafa36></span><!--]--></span></span></button></div><!----><div class="VPFlyout VPNavBarExtra extra" data-v-d2a336f3 data-v-75b97eb1 data-v-ed271fd4><button type="button" class="button" aria-haspopup="true" aria-expanded="false" aria-label="extra navigation" data-v-ed271fd4><span class="vpi-more-horizontal icon" data-v-ed271fd4></span></button><div class="menu" data-v-ed271fd4><div class="VPMenu" data-v-ed271fd4 data-v-505057d2><!----><!--[--><!--[--><!----><div class="group" data-v-75b97eb1><div class="item appearance" data-v-75b97eb1><p class="label" data-v-75b97eb1>Appearance</p><div class="appearance-action" data-v-75b97eb1><button class="VPSwitch VPSwitchAppearance" type="button" role="switch" title aria-checked="false" data-v-75b97eb1 data-v-deaafa36 data-v-88da49d2><span class="check" data-v-88da49d2><span class="icon" data-v-88da49d2><!--[--><span class="vpi-sun sun" data-v-deaafa36></span><span class="vpi-moon moon" data-v-deaafa36></span><!--]--></span></span></button></div></div></div><!----><!--]--><!--]--></div></div></div><!--[--><!--]--><button type="button" class="VPNavBarHamburger hamburger" aria-label="mobile navigation" aria-expanded="false" aria-controls="VPNavScreen" data-v-d2a336f3 data-v-966f1ac7><span class="container" data-v-966f1ac7><span class="top" data-v-966f1ac7></span><span class="middle" data-v-966f1ac7></span><span class="bottom" data-v-966f1ac7></span></span></button></div></div></div></div><div class="divider" data-v-d2a336f3><div class="divider-line" data-v-d2a336f3></div></div></div><!----></header><div class="VPLocalNav has-sidebar empty" data-v-282c430e data-v-f96f8409><div class="container" data-v-f96f8409><button class="menu" aria-expanded="false" aria-controls="VPSidebarNav" data-v-f96f8409><span class="vpi-align-left menu-icon" data-v-f96f8409></span><span class="menu-text" data-v-f96f8409>Menu</span></button><div class="VPLocalNavOutlineDropdown" style="--vp-vh:0px;" data-v-f96f8409 data-v-1bc67f4b><button data-v-1bc67f4b>Return to top</button><!----></div></div></div><aside class="VPSidebar" data-v-282c430e data-v-3d1258d2><div class="curtain" data-v-3d1258d2></div><nav class="nav" id="VPSidebarNav" aria-labelledby="sidebar-aria-label" tabindex="-1" data-v-3d1258d2><span class="visually-hidden" id="sidebar-aria-label" data-v-3d1258d2> Sidebar Navigation </span><!--[--><!--]--><!--[--><div class="no-transition group" data-v-f2306e18><section class="VPSidebarItem level-0" data-v-f2306e18 data-v-61fcdc7c><div class="item" role="button" tabindex="0" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><h2 class="text" data-v-61fcdc7c>Getting started</h2><!----></div><div class="items" data-v-61fcdc7c><!--[--><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Overview</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/quickstart.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Quickstart</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/scaffolding-agents.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Scaffold an agent with Cursor</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/guides/convert-automation.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Convert a Cursor Automation</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/concepts.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Concepts</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-f2306e18><section class="VPSidebarItem level-0" data-v-f2306e18 data-v-61fcdc7c><div class="item" role="button" tabindex="0" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><h2 class="text" data-v-61fcdc7c>Templates</h2><!----></div><div class="items" data-v-61fcdc7c><!--[--><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/templates/demo.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Demo agent</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/templates/security-reviewer.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Security reviewer</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/templates/agentic-owners.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Agentic Owners</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/templates/pr-autofixer.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>PR autofixer</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/templates/triage.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Triage agent</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-f2306e18><section class="VPSidebarItem level-0" data-v-f2306e18 data-v-61fcdc7c><div class="item" role="button" tabindex="0" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><h2 class="text" data-v-61fcdc7c>Improving agents</h2><!----></div><div class="items" data-v-61fcdc7c><!--[--><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/building-with-agents.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Building agents with agents</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/evals.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Evals</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/ab.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Live A/B metrics</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/storage.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Storage</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/hillclimbing.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Hillclimbing</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-f2306e18><section class="VPSidebarItem level-0" data-v-f2306e18 data-v-61fcdc7c><div class="item" role="button" tabindex="0" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><h2 class="text" data-v-61fcdc7c>Guides</h2><!----></div><div class="items" data-v-61fcdc7c><!--[--><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/guides/webhooks.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Webhooks & custom channels</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/guides/github.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>GitHub</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/guides/slack.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Slack</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/guides/human-in-the-loop.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Human-in-the-loop approvals</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/guides/agent-to-agent.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Agent-to-agent</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/guides/mcp-oauth.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Host MCP OAuth</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/guides/cloud-runtime.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Cloud runtime</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/guides/opentelemetry.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>OpenTelemetry</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-f2306e18><section class="VPSidebarItem level-0 has-active" data-v-f2306e18 data-v-61fcdc7c><div class="item" role="button" tabindex="0" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><h2 class="text" data-v-61fcdc7c>Example agents</h2><!----></div><div class="items" data-v-61fcdc7c><!--[--><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/example-agents/" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Choose an example</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/example-agents/weather-agent.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Weather agent</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/example-agents/slack-agent.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Slack agent</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/example-agents/concierge.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Concierge</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/example-agents/benny.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Playbook router</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/example-agents/oncall.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Alert investigator</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/example-agents/bugbot.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>PR evidence reviewer</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/example-agents/approval-buddy.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Approval Buddy</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/example-agents/security-reviewer.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Security Reviewer</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/example-agents/knowledge-base.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Knowledge base</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/example-agents/codebase-wiki.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Codebase wiki</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/example-agents/codeowners-review.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Codeowners review</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-f2306e18><section class="VPSidebarItem level-0" data-v-f2306e18 data-v-61fcdc7c><div class="item" role="button" tabindex="0" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><h2 class="text" data-v-61fcdc7c>Operating</h2><!----></div><div class="items" data-v-61fcdc7c><!--[--><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/deployment.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Deployment</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/troubleshooting.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Fix common problems</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-f2306e18><section class="VPSidebarItem level-0 collapsible" data-v-f2306e18 data-v-61fcdc7c><div class="item" role="button" tabindex="0" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><h2 class="text" data-v-61fcdc7c>Reference</h2><div class="caret" role="button" aria-label="toggle section" tabindex="0" data-v-61fcdc7c><span class="vpi-chevron-right caret-icon" data-v-61fcdc7c></span></div></div><div class="items" data-v-61fcdc7c><!--[--><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/reference/project-layout.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Project layout</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/reference/agent-config.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Agent config</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/reference/instructions.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Instructions</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/reference/tools.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Tools</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/reference/prompt.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>prompt</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/reference/skills.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Skills</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/reference/connections.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>MCP Connections</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/reference/subagents.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Subagents</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/reference/channels.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Channels</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/reference/schedules.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Schedules & reminders</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/reference/hooks.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Hooks</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/reference/artifacts.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Artifacts</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/reference/sessions.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Sessions & streaming</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/reference/playground.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>Playground</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/reference/cli.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>CLI</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-61fcdc7c><div class="item" data-v-61fcdc7c><div class="indicator" data-v-61fcdc7c></div><a class="VPLink link link" href="/docs/reference/http-api.html" data-v-61fcdc7c><!--[--><p class="text" data-v-61fcdc7c>HTTP API</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><!--]--><!--[--><!--]--></nav></aside><div class="VPContent has-sidebar" id="VPContent" data-v-282c430e data-v-3b8d81ee><div class="VPDoc has-sidebar has-aside" data-v-3b8d81ee data-v-2e506a34><!--[--><!--]--><div class="container" data-v-2e506a34><div class="aside" data-v-2e506a34><div class="aside-curtain" data-v-2e506a34></div><div class="aside-container" data-v-2e506a34><div class="aside-content" data-v-2e506a34><div class="VPDocAside" data-v-2e506a34 data-v-b485a79e><!--[--><!--]--><!--[--><!--]--><nav aria-labelledby="doc-outline-aria-label" class="VPDocAsideOutline" data-v-b485a79e data-v-cc6b345c><div class="content" data-v-cc6b345c><div class="outline-marker" data-v-cc6b345c></div><div aria-level="2" class="outline-title" id="doc-outline-aria-label" role="heading" data-v-cc6b345c>On this page</div><ul class="VPDocOutlineItem root" data-v-cc6b345c data-v-4f0215bf><!--[--><!--]--></ul></div></nav><!--[--><!--]--><div class="spacer" data-v-b485a79e></div><!--[--><!--]--><!----><!--[--><!--]--><!--[--><!--]--></div></div></div></div><div class="content" data-v-2e506a34><div class="content-container" data-v-2e506a34><!--[--><!--]--><main class="main" data-v-2e506a34><div style="position:relative;" class="vp-doc _docs_example-agents_approval-buddy" data-v-2e506a34><div><h1 id="keep-pr-approval-policy-deterministic-with-approval-buddy" tabindex="-1">Keep PR approval policy deterministic with Approval Buddy <a class="header-anchor" href="#keep-pr-approval-policy-deterministic-with-approval-buddy" aria-label="Permalink to &quot;Keep PR approval policy deterministic with Approval Buddy&quot;">​</a></h1><p>Approval Buddy approves eligible pull requests from a fixed roster and declines every other request. GitHub still blocks self-approval when the stamp identity authored the PR. Code decides eligibility. The model prepares evidence, runs two specialist reviews, and passes their findings to the approval tool without changing the policy decision.</p><p>Use this example when an agent can make a judgment inside a workflow, but authorization and the final side effect must stay in deterministic code.</p><p><a href="./../../examples/approval-buddy/">Browse the Approval Buddy source.</a></p><h2 id="keep-approval-policy-in-code" tabindex="-1">Keep approval policy in code <a class="header-anchor" href="#keep-approval-policy-in-code" aria-label="Permalink to &quot;Keep approval policy in code&quot;">​</a></h2><p>Approval Buddy draws three hard boundaries:</p><ul><li><code>prepare_review</code> and <code>approve_pr</code> re-read the live PR and apply the same eligibility rules.</li><li>Two subagents inspect prepared evidence, but their findings never grant or block approval.</li><li>Only <code>approve_pr</code> posts the GitHub review.</li></ul><p>A spoofed webhook, Slack message, or model claim can&#39;t add someone to the buddy roster. The mutating tool checks the source of truth immediately before it acts.</p><h2 id="follow-the-intended-stamp-flow" tabindex="-1">Follow the intended stamp flow <a class="header-anchor" href="#follow-the-intended-stamp-flow" aria-label="Permalink to &quot;Follow the intended stamp flow&quot;">​</a></h2><p>The root instructions ask the model to run this sequence for a qualifying PR:</p><ol><li>A non-draft <code>pull_request</code> event arrives with action <code>opened</code>, <code>reopened</code>, or <code>ready_for_review</code>.</li><li>The GitHub channel checks its repository allowlist and starts a session.</li><li><code>turn.started</code> posts a pending commit status.</li><li>The model calls <code>prepare_review</code>.</li><li>Host code fetches the live PR. It checks the author, open state, merged state, and draft state.</li><li>A qualifying PR gets <code>pr/MANIFEST.md</code>, <code>pr/meta.json</code>, and <code>pr/diff.patch</code> in the session workspace. Diffs above 2,000,000 characters are truncated and marked in metadata.</li><li>The model calls both review subagents through the built-in <code>task</code> tool.</li><li>It concatenates their contracted replies and calls <code>approve_pr</code>.</li><li><code>approve_pr</code> re-runs eligibility, posts an <code>APPROVE</code> review, and returns the outcome.</li><li>The channel posts a final commit status. A self-approval block also gets a short timeline comment because no approval review can appear.</li></ol><p>Ineligible PRs skip evidence and subagents. The model still calls <code>approve_pr</code> so the deterministic tool returns the formal decline reason.</p><p>Steps 4 through 9 are prompt-driven. The channel doesn&#39;t enforce tool order or prove both subagents ran, and <code>approve_pr</code> accepts missing findings. A failed turn clears the pending status with a green non-blocking result without approving the PR.</p><h2 id="map-the-framework-features" tabindex="-1">Map the framework features <a class="header-anchor" href="#map-the-framework-features" aria-label="Permalink to &quot;Map the framework features&quot;">​</a></h2><table tabindex="0"><thead><tr><th>Capability</th><th>Source</th><th>Role</th></tr></thead><tbody><tr><td>Root agent and policy prompt</td><td><a href="../../examples/approval-buddy/agent/agent.ts"><code>agent/agent.ts</code></a>, <a href="./../../examples/approval-buddy/agent/instructions.html"><code>agent/instructions.md</code></a></td><td>Configure the local agent and describe orchestration order.</td></tr><tr><td>GitHub channel</td><td><a href="../../examples/approval-buddy/agent/channels/github.ts"><code>agent/channels/github.ts</code></a></td><td>Filter wakes, lease GitHub access, and publish status events.</td></tr><tr><td>Slack channel</td><td><a href="../../examples/approval-buddy/agent/channels/slack.ts"><code>agent/channels/slack.ts</code></a></td><td>Accept approval-bot stamp and qualification requests.</td></tr><tr><td>Server tools</td><td><a href="./../../examples/approval-buddy/agent/tools/"><code>agent/tools/</code></a></td><td>Prepare evidence, approve, list buddies, and search GIFs.</td></tr><tr><td>Deterministic policy</td><td><a href="../../examples/approval-buddy/agent/lib/approve.ts"><code>agent/lib/approve.ts</code></a>, <a href="../../examples/approval-buddy/agent/lib/buddies.ts"><code>agent/lib/buddies.ts</code></a></td><td>Own the roster and live eligibility checks.</td></tr><tr><td>Review subagents</td><td><a href="./../../examples/approval-buddy/agent/subagents/"><code>agent/subagents/</code></a></td><td>Run deep audit and code-quality passes over the same evidence.</td></tr><tr><td>Storage</td><td><a href="../../examples/approval-buddy/agent/storage.ts"><code>agent/storage.ts</code></a></td><td>Persist sessions and events with <code>cursorHostedStorage</code>. See <a href="./../storage.html">Storage</a>.</td></tr><tr><td>Live A/B experiment</td><td><a href="../../examples/approval-buddy/agent/ab.ts"><code>agent/ab.ts</code></a></td><td>Compare baseline responses with a concise, presentation-only treatment (<code>concise-results</code>).</td></tr><tr><td>Evals and unit tests</td><td><a href="./../../examples/approval-buddy/evals/"><code>evals/</code></a>, <a href="./../../examples/approval-buddy/agent/lib/"><code>agent/lib/</code></a></td><td>Protect routing, output contracts, policy, and GitHub behavior.</td></tr></tbody></table><p>There are no authored skills, MCP connections, schedules, reminders, hooks, sandbox seeds, or tool approvals.</p><h2 id="prepare-credentials" tabindex="-1">Prepare credentials <a class="header-anchor" href="#prepare-credentials" aria-label="Permalink to &quot;Prepare credentials&quot;">​</a></h2><p>You need:</p><ul><li>Node 22.13 or newer.</li><li>An agent-runtime credential.</li><li>GitHub access to read PRs, post reviews, create commit statuses, and post the self-approval visibility comment.</li></ul><p>Optional GIF selection uses:</p><ul><li><code>GIPHY_API_KEY</code> or <code>APPROVAL_BUDDY_GIPHY_API_KEY</code>,</li><li><code>APPROVAL_BUDDY_STAMP_GIF</code>, or</li><li>severity-specific <code>APPROVAL_BUDDY_STAMP_GIF_&lt;LEVEL&gt;</code> variables.</li></ul><p>If you enable Giphy in a hosted copy, declare its secret and <code>api.giphy.com</code> egress.</p><h2 id="validate-without-approving-a-pr" tabindex="-1">Validate without approving a PR <a class="header-anchor" href="#validate-without-approving-a-pr" aria-label="Permalink to &quot;Validate without approving a PR&quot;">​</a></h2><div class="language-bash vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">bash</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">agent-sdk</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> validate</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> --dir</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> examples/approval-buddy</span></span>
-<span class="line"><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">agent-sdk</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> info</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> --dir</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> examples/approval-buddy</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> --json</span></span></code></pre></div><p>List the deterministic roster:</p><div class="language-bash vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">bash</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">agent-sdk</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> call</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> list_buddies</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> \</span></span>
-<span class="line"><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">  --dir</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> examples/approval-buddy</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> \</span></span>
-<span class="line"><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">  --input</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> &#39;{}&#39;</span></span></code></pre></div><p>Set a known merged PR, then run the read-only precheck:</p><div class="language-bash vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">bash</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">MERGED_PR_URL</span><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">=</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">https://github.com/your-org/your-repo/pull/123</span></span>
-<span class="line"><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">agent-sdk</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> call</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> prepare_review</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> \</span></span>
-<span class="line"><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">  --dir</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> examples/approval-buddy</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> \</span></span>
-<span class="line"><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">  --input</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> &quot;{</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">\&quot;</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">prUrl</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">\&quot;</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">:</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">\&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">$MERGED_PR_URL</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">\&quot;</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">}&quot;</span></span></code></pre></div><p>The result should decline because the PR is no longer open. <code>prepare_review</code> never posts an approval.</p><div class="caution custom-block github-alert"><p class="custom-block-title">CAUTION</p><p>Don&#39;t use <code>agent-sdk call approve_pr</code> as a smoke test. The tool has no <code>needsApproval</code> gate and posts a real GitHub review when the PR qualifies.</p></div><h2 id="see-why-preparation-is-separate" tabindex="-1">See why preparation is separate <a class="header-anchor" href="#see-why-preparation-is-separate" aria-label="Permalink to &quot;See why preparation is separate&quot;">​</a></h2><p><code>prepare_review</code> is read-only. It checks policy before fetching a large diff, so declined requests don&#39;t spend review-agent work.</p><p>Direct calls return the evidence file map because their scratch workspace is deleted after the call. In-session calls write the tree to <code>ctx.workspaceDir</code>, where both subagents can read it.</p><p><code>approve_pr</code> repeats the live check instead of trusting preparation. A PR can close, merge, become a draft, or change author-related context between the two steps. Revalidation keeps the final write bound to current state.</p><p>This is a reusable two-tool pattern:</p><ul><li>a read-only tool prepares and explains the decision,</li><li>a mutating tool repeats policy at the side-effect boundary.</li></ul><h2 id="fan-out-two-review-contracts" tabindex="-1">Fan out two review contracts <a class="header-anchor" href="#fan-out-two-review-contracts" aria-label="Permalink to &quot;Fan out two review contracts&quot;">​</a></h2><p>The two discovered subagents have different contracts:</p><ul><li>The security reviewer reports bugs, breaking changes, and security findings with <code>High</code>, <code>Medium</code>, or <code>Low</code> tags.</li><li>The code-quality reviewer reports maintainability and structure concerns with <code>Blocker</code>, <code>Major</code>, or <code>Minor</code> tags.</li></ul><p>The parent calls both through the harness <code>task</code> tool. They inherit the root agent&#39;s execution surface and read the same <code>pr/</code> workspace. The prompt asks the parent not to rewrite either reply. The review body trims the combined text and caps it at 16,000 characters.</p><p>Findings are informational. A high-severity finding doesn&#39;t veto the stamp. That policy is explicit in the root instructions and approval code.</p><h2 id="trace-github-channel-behavior" tabindex="-1">Trace GitHub channel behavior <a class="header-anchor" href="#trace-github-channel-behavior" aria-label="Permalink to &quot;Trace GitHub channel behavior&quot;">​</a></h2><p>The channel uses <code>githubChannel</code> with:</p><ul><li>a configured repository allowlist on the account-linked GitHub transport,</li><li>a second optional <code>APPROVAL_BUDDY_REPOS</code> wake filter,</li><li><code>deliverReplies: false</code>,</li><li>progress reactions disabled, and</li><li>event handlers for turn start, <code>approve_pr</code> results, and failed turns.</li></ul><p>The source requests <code>contents-write</code>, even though the documented workflow posts reviews, statuses, and comments. When adapting the example, start with <code>pr-write</code> and opt up only if a tool must push code.</p><p>Every terminal status is green by design. Declines and crashed turns are informational, not merge-blocking. This is a product decision in the example, not an Agent SDK default.</p><p>A successful turn that never calls <code>approve_pr</code> leaves the pending status in place. The channel clears it on <code>approve_pr</code> results and <code>turn.failed</code>, but has no <code>turn.completed</code> fallback.</p><p><code>github replay</code> reaches the same channel and can post a real approval, status, or comment. Use replay only against a repository and PR created for this test.</p><h2 id="use-slack-for-explicit-requests" tabindex="-1">Use Slack for explicit requests <a class="header-anchor" href="#use-slack-for-explicit-requests" aria-label="Permalink to &quot;Use Slack for explicit requests&quot;">​</a></h2><p>Start the dev server:</p><div class="language-bash vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">bash</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">agent-sdk</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> dev</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> examples/approval-buddy</span></span></code></pre></div><p>Then ask through the signed-in account-linked Slack connection:</p><blockquote><p>Would this PR qualify for a stamp?</p></blockquote><p>The instructions route qualification questions to <code>prepare_review</code> only. A stamp request runs the complete flow and may approve the PR.</p><p>This channel uses the account-linked transport instead of a dedicated Socket Mode app.</p><h2 id="see-how-durable-storage-fits" tabindex="-1">See how durable storage fits <a class="header-anchor" href="#see-how-durable-storage-fits" aria-label="Permalink to &quot;See how durable storage fits&quot;">​</a></h2><p><code>defineStorage</code> replaces the default local session store with a shared, durable key-value adapter. Approval Buddy chooses:</p><ul><li>a 15-second write debounce,</li><li>startup restoration for up to 200 sessions, and</li><li>a 14-day restore window.</li></ul><p>That policy fits long-lived Slack threads and a small webhook fleet. The security reviewer uses the same adapter with lazy restore, which fits its shorter sessions.</p><h2 id="run-the-regression-suite" tabindex="-1">Run the regression suite <a class="header-anchor" href="#run-the-regression-suite" aria-label="Permalink to &quot;Run the regression suite&quot;">​</a></h2><p>List the four eval cases:</p><div class="language-bash vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">bash</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">agent-sdk</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> eval</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> --dir</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> examples/approval-buddy</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> --list</span></span></code></pre></div><p>The suite covers:</p><ul><li>buddy-list routing,</li><li>declining a merged PR,</li><li>using only <code>prepare_review</code> for a qualification question, and</li><li>the combined findings headings and severity format over seeded evidence.</li></ul><p>Run the safe qualification case:</p><div class="language-bash vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">bash</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">agent-sdk</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> eval</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> \</span></span>
-<span class="line"><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">  --dir</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> examples/approval-buddy</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> \</span></span>
-<span class="line"><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">  qualify/merged-pr-question</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> \</span></span>
-<span class="line"><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">  --json</span></span></code></pre></div><p>The qualification case reads a live merged PR. The seeded format case shown by <code>--list</code> uses a planted auth-bypass diff and checks for a <code>task</code> call, both headings, and severity tags. It doesn&#39;t prove both named subagents ran or whether their output reached <code>approve_pr</code>. Unit tests under <code>agent/lib/</code> cover policy, self-approval handling, evidence limits, status mapping, GIF selection, and severity parsing.</p><h2 id="reuse-the-policy-boundary" tabindex="-1">Reuse the policy boundary <a class="header-anchor" href="#reuse-the-policy-boundary" aria-label="Permalink to &quot;Reuse the policy boundary&quot;">​</a></h2><p>Keep these properties when you replace the buddy policy:</p><ol><li>Put authorization in typed code.</li><li>Fetch the source of truth inside both prepare and mutate steps.</li><li>Give the model evidence only after the request qualifies.</li><li>Treat specialist findings as data, not authority.</li><li>Keep the core domain mutation in one named tool. Treat channel status and visibility writes as separate, audited effects.</li><li>Add a human approval gate if your policy still needs operator consent.</li><li>Test read-only routing separately from mutation.</li></ol><h2 id="where-to-go-next" tabindex="-1">Where to go next <a class="header-anchor" href="#where-to-go-next" aria-label="Permalink to &quot;Where to go next&quot;">​</a></h2><ul><li><a href="./../guides/github.html">GitHub</a></li><li><a href="./../reference/tools.html">Tools</a></li><li><a href="./../reference/subagents.html">Subagents</a></li><li><a href="./../storage.html">Storage</a></li><li><a href="./../guides/slack.html">Slack</a></li><li><a href="./../evals.html">Evals</a></li></ul></div></div></main><footer class="VPDocFooter" data-v-2e506a34 data-v-574f06fa><!--[--><!--]--><!----><nav class="prev-next" aria-labelledby="doc-footer-aria-label" data-v-574f06fa><span class="visually-hidden" id="doc-footer-aria-label" data-v-574f06fa>Pager</span><div class="pager" data-v-574f06fa><a class="VPLink link pager-link prev" href="/docs/example-agents/bugbot.html" data-v-574f06fa><!--[--><span class="desc" data-v-574f06fa>Previous page</span><span class="title" data-v-574f06fa>PR evidence reviewer</span><!--]--></a></div><div class="pager" data-v-574f06fa><a class="VPLink link pager-link next" href="/docs/example-agents/security-reviewer.html" data-v-574f06fa><!--[--><span class="desc" data-v-574f06fa>Next page</span><span class="title" data-v-574f06fa>Security Reviewer</span><!--]--></a></div></nav></footer><!--[--><!--]--></div></div></div><!--[--><!--]--></div></div><!----><!--[--><!--]--></div></div>
-    <script>window.__VP_HASH_MAP__=JSON.parse("{\"ab.md\":\"DJo5r4R-\",\"building-with-agents.md\":\"DI4mEzlt\",\"concepts.md\":\"F6AiPorA\",\"deployment.md\":\"DoLFAzfm\",\"evals.md\":\"lfJoEVc8\",\"example-agents_approval-buddy.md\":\"DmezILPg\",\"example-agents_benny.md\":\"B0kwY7D_\",\"example-agents_bugbot.md\":\"BRGMi9O2\",\"example-agents_codebase-wiki.md\":\"BBNw9Ekr\",\"example-agents_codeowners-review.md\":\"Bfta-lBU\",\"example-agents_concierge.md\":\"BzB2b20R\",\"example-agents_index.md\":\"ChBp0AX6\",\"example-agents_knowledge-base.md\":\"CrA85ig-\",\"example-agents_oncall.md\":\"DK4XkYTd\",\"example-agents_security-reviewer.md\":\"74pPpWYj\",\"example-agents_slack-agent.md\":\"D7Kdj5BV\",\"example-agents_weather-agent.md\":\"CaGpmw3Y\",\"guides_agent-to-agent.md\":\"B3JIaAqz\",\"guides_cloud-runtime.md\":\"BnvjPiia\",\"guides_convert-automation.md\":\"Bboisykk\",\"guides_github.md\":\"DqJhuaN1\",\"guides_human-in-the-loop.md\":\"By1G2T3_\",\"guides_mcp-oauth.md\":\"CJvrXtkN\",\"guides_opentelemetry.md\":\"bmPmkvJu\",\"guides_slack.md\":\"mqeNKs84\",\"guides_webhooks.md\":\"DKdA43Qm\",\"hillclimbing.md\":\"DhESf3OO\",\"index.md\":\"B-lVR4wT\",\"quickstart.md\":\"BrmfrrIr\",\"reference_agent-config.md\":\"Cp_x38Nl\",\"reference_artifacts.md\":\"Dior32Qw\",\"reference_channels.md\":\"Cd2f2iyV\",\"reference_cli.md\":\"D9KESDsD\",\"reference_connections.md\":\"DB6SsN6U\",\"reference_hooks.md\":\"BxN87gCw\",\"reference_http-api.md\":\"C68BERYr\",\"reference_instructions.md\":\"CR7XSsGk\",\"reference_playground.md\":\"DnX5nL-B\",\"reference_project-layout.md\":\"WN9nwJht\",\"reference_prompt.md\":\"DnaD5dNK\",\"reference_schedules.md\":\"DI_JrHgq\",\"reference_sessions.md\":\"D0mIh4KK\",\"reference_skills.md\":\"BFW9retM\",\"reference_subagents.md\":\"Xoav0AII\",\"reference_tools.md\":\"DuKvkYWG\",\"scaffolding-agents.md\":\"D7UUkWw0\",\"storage.md\":\"BOHeqk2M\",\"templates_agentic-owners.md\":\"DqtPdm6f\",\"templates_demo.md\":\"DhFcWN6j\",\"templates_pr-autofixer.md\":\"R4K_qytS\",\"templates_security-reviewer.md\":\"ByFyRta2\",\"templates_triage.md\":\"CVlpctKS\",\"troubleshooting.md\":\"vCWwvqcJ\"}");window.__VP_SITE_DATA__=JSON.parse("{\"lang\":\"en-US\",\"dir\":\"ltr\",\"title\":\"Agent SDK\",\"description\":\"Filesystem-first framework for building and serving Cursor agents.\",\"base\":\"/docs/\",\"head\":[],\"router\":{\"prefetchLinks\":true},\"appearance\":true,\"themeConfig\":{\"sdkVersion\":\"0.1.93\",\"nav\":[{\"text\":\"Quickstart\",\"link\":\"/quickstart\"},{\"text\":\"Guides\",\"link\":\"/guides/webhooks\"},{\"text\":\"Examples\",\"link\":\"/example-agents/\"},{\"text\":\"Reference\",\"link\":\"/reference/project-layout\"}],\"search\":{\"provider\":\"local\"},\"outline\":{\"level\":[2,3]},\"sidebar\":[{\"text\":\"Getting started\",\"items\":[{\"text\":\"Overview\",\"link\":\"/\"},{\"text\":\"Quickstart\",\"link\":\"/quickstart\"},{\"text\":\"Scaffold an agent with Cursor\",\"link\":\"/scaffolding-agents\"},{\"text\":\"Convert a Cursor Automation\",\"link\":\"/guides/convert-automation\"},{\"text\":\"Concepts\",\"link\":\"/concepts\"}]},{\"text\":\"Templates\",\"items\":[{\"text\":\"Demo agent\",\"link\":\"/templates/demo\"},{\"text\":\"Security reviewer\",\"link\":\"/templates/security-reviewer\"},{\"text\":\"Agentic Owners\",\"link\":\"/templates/agentic-owners\"},{\"text\":\"PR autofixer\",\"link\":\"/templates/pr-autofixer\"},{\"text\":\"Triage agent\",\"link\":\"/templates/triage\"}]},{\"text\":\"Improving agents\",\"items\":[{\"text\":\"Building agents with agents\",\"link\":\"/building-with-agents\"},{\"text\":\"Evals\",\"link\":\"/evals\"},{\"text\":\"Live A/B metrics\",\"link\":\"/ab\"},{\"text\":\"Storage\",\"link\":\"/storage\"},{\"text\":\"Hillclimbing\",\"link\":\"/hillclimbing\"}]},{\"text\":\"Guides\",\"items\":[{\"text\":\"Webhooks & custom channels\",\"link\":\"/guides/webhooks\"},{\"text\":\"GitHub\",\"link\":\"/guides/github\"},{\"text\":\"Slack\",\"link\":\"/guides/slack\"},{\"text\":\"Human-in-the-loop approvals\",\"link\":\"/guides/human-in-the-loop\"},{\"text\":\"Agent-to-agent\",\"link\":\"/guides/agent-to-agent\"},{\"text\":\"Host MCP OAuth\",\"link\":\"/guides/mcp-oauth\"},{\"text\":\"Cloud runtime\",\"link\":\"/guides/cloud-runtime\"},{\"text\":\"OpenTelemetry\",\"link\":\"/guides/opentelemetry\"}]},{\"text\":\"Example agents\",\"items\":[{\"text\":\"Choose an example\",\"link\":\"/example-agents/\"},{\"text\":\"Weather agent\",\"link\":\"/example-agents/weather-agent\"},{\"text\":\"Slack agent\",\"link\":\"/example-agents/slack-agent\"},{\"text\":\"Concierge\",\"link\":\"/example-agents/concierge\"},{\"text\":\"Playbook router\",\"link\":\"/example-agents/benny\"},{\"text\":\"Alert investigator\",\"link\":\"/example-agents/oncall\"},{\"text\":\"PR evidence reviewer\",\"link\":\"/example-agents/bugbot\"},{\"text\":\"Approval Buddy\",\"link\":\"/example-agents/approval-buddy\"},{\"text\":\"Security Reviewer\",\"link\":\"/example-agents/security-reviewer\"},{\"text\":\"Knowledge base\",\"link\":\"/example-agents/knowledge-base\"},{\"text\":\"Codebase wiki\",\"link\":\"/example-agents/codebase-wiki\"},{\"text\":\"Codeowners review\",\"link\":\"/example-agents/codeowners-review\"}]},{\"text\":\"Operating\",\"items\":[{\"text\":\"Deployment\",\"link\":\"/deployment\"},{\"text\":\"Fix common problems\",\"link\":\"/troubleshooting\"}]},{\"text\":\"Reference\",\"collapsed\":false,\"items\":[{\"text\":\"Project layout\",\"link\":\"/reference/project-layout\"},{\"text\":\"Agent config\",\"link\":\"/reference/agent-config\"},{\"text\":\"Instructions\",\"link\":\"/reference/instructions\"},{\"text\":\"Tools\",\"link\":\"/reference/tools\"},{\"text\":\"prompt\",\"link\":\"/reference/prompt\"},{\"text\":\"Skills\",\"link\":\"/reference/skills\"},{\"text\":\"MCP Connections\",\"link\":\"/reference/connections\"},{\"text\":\"Subagents\",\"link\":\"/reference/subagents\"},{\"text\":\"Channels\",\"link\":\"/reference/channels\"},{\"text\":\"Schedules & reminders\",\"link\":\"/reference/schedules\"},{\"text\":\"Hooks\",\"link\":\"/reference/hooks\"},{\"text\":\"Artifacts\",\"link\":\"/reference/artifacts\"},{\"text\":\"Sessions & streaming\",\"link\":\"/reference/sessions\"},{\"text\":\"Playground\",\"link\":\"/reference/playground\"},{\"text\":\"CLI\",\"link\":\"/reference/cli\"},{\"text\":\"HTTP API\",\"link\":\"/reference/http-api\"}]}]},\"locales\":{},\"scrollOffset\":134,\"cleanUrls\":false}");</script>
-    
-  </body>
-</html>

package/dist/docs/example-agents/approval-buddy.md

@@ -1,266 +0,0 @@
-# Keep PR approval policy deterministic with Approval Buddy
-
-Approval Buddy approves eligible pull requests from a fixed roster and
-declines every other request. GitHub still blocks self-approval when the stamp
-identity authored the PR. Code decides eligibility. The model prepares
-evidence, runs two specialist reviews, and passes their findings to the
-approval tool without changing the policy decision.
-
-Use this example when an agent can make a judgment inside a workflow, but
-authorization and the final side effect must stay in deterministic code.
-
-[Browse the Approval Buddy source.](https://github.com/cursor/cursor/tree/main/packages/agent-serve/examples/approval-buddy/)
-
-## Keep approval policy in code
-
-Approval Buddy draws three hard boundaries:
-
-- `prepare_review` and `approve_pr` re-read the live PR and apply the same
-  eligibility rules.
-- Two subagents inspect prepared evidence, but their findings never grant or
-  block approval.
-- Only `approve_pr` posts the GitHub review.
-
-A spoofed webhook, Slack message, or model claim can't add someone to the
-buddy roster. The mutating tool checks the source of truth immediately before it
-acts.
-
-## Follow the intended stamp flow
-
-The root instructions ask the model to run this sequence for a qualifying PR:
-
-1. A non-draft `pull_request` event arrives with action `opened`, `reopened`,
-   or `ready_for_review`.
-2. The GitHub channel checks its repository allowlist and starts a session.
-3. `turn.started` posts a pending commit status.
-4. The model calls `prepare_review`.
-5. Host code fetches the live PR. It checks the author, open state, merged
-   state, and draft state.
-6. A qualifying PR gets `pr/MANIFEST.md`, `pr/meta.json`, and
-   `pr/diff.patch` in the session workspace. Diffs above 2,000,000
-   characters are truncated and marked in metadata.
-7. The model calls both review subagents through the built-in `task` tool.
-8. It concatenates their contracted replies and calls `approve_pr`.
-9. `approve_pr` re-runs eligibility, posts an `APPROVE` review, and returns
-   the outcome.
-10. The channel posts a final commit status. A self-approval block also gets
-    a short timeline comment because no approval review can appear.
-
-Ineligible PRs skip evidence and subagents. The model still calls
-`approve_pr` so the deterministic tool returns the formal decline reason.
-
-Steps 4 through 9 are prompt-driven. The channel doesn't enforce tool order
-or prove both subagents ran, and `approve_pr` accepts missing findings. A
-failed turn clears the pending status with a green non-blocking result without
-approving the PR.
-
-## Map the framework features
-
-| Capability | Source | Role |
-| --- | --- | --- |
-| Root agent and policy prompt | [`agent/agent.ts`](https://github.com/cursor/cursor/blob/main/packages/agent-serve/examples/approval-buddy/agent/agent.ts), [`agent/instructions.md`](https://github.com/cursor/cursor/blob/main/packages/agent-serve/examples/approval-buddy/agent/instructions.md) | Configure the local agent and describe orchestration order. |
-| GitHub channel | [`agent/channels/github.ts`](https://github.com/cursor/cursor/blob/main/packages/agent-serve/examples/approval-buddy/agent/channels/github.ts) | Filter wakes, lease GitHub access, and publish status events. |
-| Slack channel | [`agent/channels/slack.ts`](https://github.com/cursor/cursor/blob/main/packages/agent-serve/examples/approval-buddy/agent/channels/slack.ts) | Accept approval-bot stamp and qualification requests. |
-| Server tools | [`agent/tools/`](https://github.com/cursor/cursor/tree/main/packages/agent-serve/examples/approval-buddy/agent/tools/) | Prepare evidence, approve, list buddies, and search GIFs. |
-| Deterministic policy | [`agent/lib/approve.ts`](https://github.com/cursor/cursor/blob/main/packages/agent-serve/examples/approval-buddy/agent/lib/approve.ts), [`agent/lib/buddies.ts`](https://github.com/cursor/cursor/blob/main/packages/agent-serve/examples/approval-buddy/agent/lib/buddies.ts) | Own the roster and live eligibility checks. |
-| Review subagents | [`agent/subagents/`](https://github.com/cursor/cursor/tree/main/packages/agent-serve/examples/approval-buddy/agent/subagents/) | Run deep audit and code-quality passes over the same evidence. |
-| Storage | [`agent/storage.ts`](https://github.com/cursor/cursor/blob/main/packages/agent-serve/examples/approval-buddy/agent/storage.ts) | Persist sessions and events with `cursorHostedStorage`. See [Storage](/docs/storage.md). |
-| Live A/B experiment | [`agent/ab.ts`](https://github.com/cursor/cursor/blob/main/packages/agent-serve/examples/approval-buddy/agent/ab.ts) | Compare baseline responses with a concise, presentation-only treatment (`concise-results`). |
-| Evals and unit tests | [`evals/`](https://github.com/cursor/cursor/tree/main/packages/agent-serve/examples/approval-buddy/evals/), [`agent/lib/`](https://github.com/cursor/cursor/tree/main/packages/agent-serve/examples/approval-buddy/agent/lib/) | Protect routing, output contracts, policy, and GitHub behavior. |
-
-There are no authored skills, MCP connections, schedules, reminders, hooks,
-sandbox seeds, or tool approvals.
-
-## Prepare credentials
-
-You need:
-
-- Node 22.13 or newer.
-- An agent-runtime credential.
-- GitHub access to read PRs, post reviews, create commit statuses, and
-  post the self-approval visibility comment.
-
-Optional GIF selection uses:
-
-- `GIPHY_API_KEY` or `APPROVAL_BUDDY_GIPHY_API_KEY`,
-- `APPROVAL_BUDDY_STAMP_GIF`, or
-- severity-specific `APPROVAL_BUDDY_STAMP_GIF_<LEVEL>` variables.
-
-If you enable Giphy in a hosted copy, declare its secret and
-`api.giphy.com` egress.
-
-## Validate without approving a PR
-
-```bash
-agent-sdk validate --dir examples/approval-buddy
-agent-sdk info --dir examples/approval-buddy --json
-```
-
-List the deterministic roster:
-
-```bash
-agent-sdk call list_buddies \
-  --dir examples/approval-buddy \
-  --input '{}'
-```
-
-Set a known merged PR, then run the read-only precheck:
-
-```bash
-MERGED_PR_URL=https://github.com/your-org/your-repo/pull/123
-agent-sdk call prepare_review \
-  --dir examples/approval-buddy \
-  --input "{\"prUrl\":\"$MERGED_PR_URL\"}"
-```
-
-The result should decline because the PR is no longer open. `prepare_review`
-never posts an approval.
-
-> [!CAUTION]
-> Don't use `agent-sdk call approve_pr` as a smoke test. The tool has no
-> `needsApproval` gate and posts a real GitHub review when the PR qualifies.
-
-## See why preparation is separate
-
-`prepare_review` is read-only. It checks policy before fetching a large diff,
-so declined requests don't spend review-agent work.
-
-Direct calls return the evidence file map because their scratch workspace is
-deleted after the call. In-session calls write the tree to
-`ctx.workspaceDir`, where both subagents can read it.
-
-`approve_pr` repeats the live check instead of trusting preparation. A PR can
-close, merge, become a draft, or change author-related context between the two
-steps. Revalidation keeps the final write bound to current state.
-
-This is a reusable two-tool pattern:
-
-- a read-only tool prepares and explains the decision,
-- a mutating tool repeats policy at the side-effect boundary.
-
-## Fan out two review contracts
-
-The two discovered subagents have different contracts:
-
-- The security reviewer reports bugs, breaking changes, and security findings
-  with `High`, `Medium`, or `Low` tags.
-- The code-quality reviewer reports maintainability and structure concerns
-  with `Blocker`, `Major`, or `Minor` tags.
-
-The parent calls both through the harness `task` tool. They inherit the root
-agent's execution surface and read the same `pr/` workspace. The prompt asks
-the parent not to rewrite either reply. The review body trims the combined
-text and caps it at 16,000 characters.
-
-Findings are informational. A high-severity finding doesn't veto the stamp.
-That policy is explicit in the root instructions and approval code.
-
-## Trace GitHub channel behavior
-
-The channel uses `githubChannel` with:
-
-- a configured repository allowlist on the account-linked GitHub transport,
-- a second optional `APPROVAL_BUDDY_REPOS` wake filter,
-- `deliverReplies: false`,
-- progress reactions disabled, and
-- event handlers for turn start, `approve_pr` results, and failed turns.
-
-The source requests `contents-write`, even though the documented workflow
-posts reviews, statuses, and comments. When adapting the example, start with
-`pr-write` and opt up only if a tool must push code.
-
-Every terminal status is green by design. Declines and crashed turns are
-informational, not merge-blocking. This is a product decision in the example,
-not an Agent SDK default.
-
-A successful turn that never calls `approve_pr` leaves the pending status in
-place. The channel clears it on `approve_pr` results and `turn.failed`, but
-has no `turn.completed` fallback.
-
-`github replay` reaches the same channel and can post a real approval, status,
-or comment. Use replay only against a repository and PR created for this
-test.
-
-## Use Slack for explicit requests
-
-Start the dev server:
-
-```bash
-agent-sdk dev examples/approval-buddy
-```
-
-Then ask through the signed-in account-linked Slack connection:
-
-> Would this PR qualify for a stamp?
-
-The instructions route qualification questions to `prepare_review` only. A
-stamp request runs the complete flow and may approve the PR.
-
-This channel uses the account-linked transport instead of a dedicated Socket
-Mode app.
-
-## See how durable storage fits
-
-`defineStorage` replaces the default local session store with a shared,
-durable key-value adapter. Approval Buddy chooses:
-
-- a 15-second write debounce,
-- startup restoration for up to 200 sessions, and
-- a 14-day restore window.
-
-That policy fits long-lived Slack threads and a small webhook fleet. The
-security reviewer uses the same adapter with lazy restore, which fits its
-shorter sessions.
-
-## Run the regression suite
-
-List the four eval cases:
-
-```bash
-agent-sdk eval --dir examples/approval-buddy --list
-```
-
-The suite covers:
-
-- buddy-list routing,
-- declining a merged PR,
-- using only `prepare_review` for a qualification question, and
-- the combined findings headings and severity format over seeded evidence.
-
-Run the safe qualification case:
-
-```bash
-agent-sdk eval \
-  --dir examples/approval-buddy \
-  qualify/merged-pr-question \
-  --json
-```
-
-The qualification case reads a live merged PR. The seeded format case shown
-by `--list` uses a planted auth-bypass diff
-and checks for a `task` call, both headings, and severity tags. It doesn't
-prove both named subagents ran or whether their output reached `approve_pr`.
-Unit tests under `agent/lib/` cover policy, self-approval handling, evidence
-limits, status mapping, GIF selection, and severity parsing.
-
-## Reuse the policy boundary
-
-Keep these properties when you replace the buddy policy:
-
-1. Put authorization in typed code.
-2. Fetch the source of truth inside both prepare and mutate steps.
-3. Give the model evidence only after the request qualifies.
-4. Treat specialist findings as data, not authority.
-5. Keep the core domain mutation in one named tool. Treat channel status and
-   visibility writes as separate, audited effects.
-6. Add a human approval gate if your policy still needs operator consent.
-7. Test read-only routing separately from mutation.
-
-## Where to go next
-
-- [GitHub](/docs/guides/github.md)
-- [Tools](/docs/reference/tools.md)
-- [Subagents](/docs/reference/subagents.md)
-- [Storage](/docs/storage.md)
-- [Slack](/docs/guides/slack.md)
-- [Evals](/docs/evals.md)